A discrete-time data-driven self-learning unmanned ship speed control system
By introducing expansion state observer and data-driven update law in the unmanned ship speed control system, the unknown dynamic terms and total system disturbance of the unmanned ship are estimated, and the slow convergence speed of control gain changes and adaptive parameters in the prior art are solved. The problem of estimation of variable control gain and system self-learning function is realized.
Patent Information
- Application Number
- CN202210272085.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-03-18
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2042-03-18
AI Technical Summary
The existing unmanned ship speed self-immune control method requires a given control gain coefficient in advance, and the changing control gain cannot be processed, resulting in limited application scope and a large amount of experimental data is required when estimating uncertainty, and the convergence speed of adaptive parameters is slow. .
A discrete time data-driven self-learning unmanned ship speed control system is designed, including speed estimation module, disturbance observer module, modelless controller module and data-driven parameter update law module, which is updated through expansion state observer and data-driven The law, estimates the unknown dynamic terms of the unmanned ship and the total system perturbation, realizes the estimation and self-learning functions of variable control gain.
Based on the estimation of the total disturbance, the control gain and its inaccurate estimation can be effectively estimated, which improves the tracking response speed of the control gain, realizes the system's complete self-learning and monitoring functions, and solves the estimation and self-authorization of the change control gain. To adapt to the problem of slow convergence of parameters.
Smart Images

Figure CN114564028B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of unmanned ship speed control, and in particular to an unmanned ship speed control system driven by discrete time data and self-learning. Background Art
[0002] The motion control of unmanned ships includes two aspects: kinematic control and dynamic control. The dynamic control of unmanned ships refers to the tracking of a given speed by adjusting the actuators to generate corresponding forces and torques. In speed control, the dynamic model of unmanned ships has internal model uncertainty and unknown marine environment disturbances. In order to improve the anti-disturbance performance of unmanned ships and keep the speed and angular velocity of unmanned ships stable in special marine environments, the unified estimation of internal and external disturbances of the system has become a key issue. Domestic and foreign scholars have carried out a lot of related research. On the basis of classical control and modern control theory, the existing dynamic control method analyzes and estimates the motion state and uncertainty of unmanned ships by introducing least squares support vector machine regression, fuzzy logic control, networked ship adaptive system identification, neural network and other methods, so as to achieve accurate tracking of a given speed in a complex environment. Among them, the disturbance estimation method in the anti-disturbance control and the parameter-based online adaptive method in the adaptive control are more successful in the application of the motion state and uncertainty estimation of unmanned ships. From the perspective of the existing speed controller design, the existing unmanned ship speed control method still has the following shortcomings:
[0003] In the existing unmanned ship speed self-disturbance rejection control method, the control system often needs to give the control gain coefficient in advance or know its nominal value. However, in the actual control process, the control gain coefficient is often unknown. In the existing unmanned ship speed self-disturbance rejection control method, the existing method can only estimate the given control gain, but cannot handle the changing control gain, so the application scope has certain limitations. Summary of the invention
[0004] The present invention provides a discrete-time data driven self-learning unmanned ship speed control system to overcome the above problems.
[0005] The present invention comprises: a speed estimation module, a disturbance observer module, a model-free controller module, and a data-driven parameter update law module;
[0006] The model-free controller module is used to calculate the forward torque τ(n);
[0007] The speed estimation module is used to calculate the unknown dynamics of the unmanned ship at the next moment according to the forward torque τ(n). Estimation of the speed of the unmanned ship at the next moment
[0008] The disturbance observer module is used to estimate the unknown dynamics of the unmanned ship at the next moment. Speed estimation of unmanned ship Calculate the total disturbance estimate of the system at the next moment
[0009] The data-driven parameter update law module is used to estimate the total disturbance of the system at the next moment. Calculate the control gain estimate for the next moment
[0010] Furthermore, the speed estimation module is:
[0011]
[0012] in, is the estimated speed of the unmanned ship at the current moment, is the estimated speed of the unmanned ship at the next moment, is the estimate of the unknown dynamics at the current moment, is the estimate of the unknown dynamics of the unmanned ship at the next moment, is the total disturbance estimate of the system at the current moment, λ 1 and λ 2 To adjust the expanded state error The introduced scalar gain parameter, T is the discrete time constant, and τ(n) is the forward torque output to push the ship forward.
[0013] Furthermore, the disturbance observer module is:
[0014]
[0015] Among them, γ 1 is the adaptive gain, is the total disturbance estimate of the system at the next moment, is the total disturbance estimate of the system at the current moment, u(n) represents the speed of the unmanned ship at the current moment, u(n+1) represents the speed of the unmanned ship at the next moment, u(n)∈R, u(n+1)∈R, T is the discrete time constant, is the total disturbance estimate of the system at the current moment, is the control gain estimate at the current moment, and τ(n) is the forward torque output to push the ship forward.
[0016] Furthermore, the calculation formula of the unmanned ship speed u(n+1) at the next moment is:
[0017] u(n+1)=u(n)+T(d+k 0 τ(n)) (3)
[0018] Among them, u(n) represents the speed of the unmanned ship at the current moment, u(n+1) represents the speed of the unmanned ship at the next moment, u(n)∈R, u(n+1)∈R; T is the discrete time constant, is the control gain, m is the mass parameter of the unmanned ship, τ(n) is the forward torque output to push the ship forward, and d is the dynamic unknown term.
[0019] Furthermore, the calculation formula of the kinetic unknown term d is:
[0020] d=m -1 X u u(n)+m -1 X u|u| u(n)|u(n)|+m -1 ω (4)
[0021] Among them, d is the unknown term of dynamics, m is the mass parameter of the unmanned ship, and X u With X u|u| is the hydrodynamic parameter, u(n) represents the speed of the unmanned ship at the current moment, u(n)∈R; ω is the external disturbance suffered by the unmanned ship.
[0022] Furthermore, the model-free controller module is:
[0023]
[0024] Among them, ω c is the model-free control gain, τ(n) is the forward torque output to push the ship forward, u(n) represents the speed of the unmanned ship at the current moment, u(n)∈R; is the control gain estimate at the current moment, is the total disturbance estimate of the system at the current moment, Estimate the unknown dynamics at the current moment.
[0025] Furthermore, the data-driven parameter update law module is:
[0026]
[0027] in, is the control gain estimate for the next moment, γ 2 and γ 3 are the adaptive gains of the data-driven parameter update law module, is the control gain estimate for the next moment; τ(n) is the forward torque output to push the ship forward, τ T(n) is the inverse matrix of the forward torque that outputs the ship forward; u(n) represents the speed of the unmanned ship at the current moment, u(n)∈R; u(n+1) represents the speed of the unmanned ship at the next moment, u(n)∈R, u(n+1)∈R; T is the discrete time constant, is the control gain estimate at the current moment, is the total disturbance estimate of the system at the current moment.
[0028] Beneficial effects:
[0029] The present invention not only solves the problem of estimating the variable control gain, but also identifies the inaccurate part of the control gain estimation based on the estimation of the total disturbance, and finally realizes the effect of the system's complete self-learning and monitoring functions by utilizing the extended state observer and the data-driven update law. BRIEF DESCRIPTION OF THE DRAWINGS
[0030] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0031] Figure 1 is a system block diagram of the present invention;
[0032] Figure 2 It is a general schematic diagram of the present invention;
[0033] Figure 3 This is the control effect diagram of the unmanned ship speed u;
[0034] Figure 4 is the observed effect diagram of the total disturbance δ of the system;
[0035] Figure 5 is the given signal diagram of the forward torque τ;
[0036] Figure 6 is the control gain k 0 Observation effect diagram. DETAILED DESCRIPTION
[0037] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0038] like Figure 1 As shown, the present invention includes: a speed estimation module, a disturbance observer module, a model-free controller module, and a data-driven parameter update law module;
[0039] The model-free controller module is used to calculate the forward torque τ(n);
[0040] Specifically, the forward torque is used as the controller of the system. Its function is to generate an unmanned ship speed closer to the expected speed through the difference between the speed of the unmanned ship at the current moment and the next moment, and use the output of the model-free controller module as the input of other modules in the system.
[0041] The speed estimation module is used to calculate the unknown dynamics of the unmanned ship at the next moment according to the forward torque τ(n). Speed estimation of unmanned ship
[0042] Specifically, the speed estimation module estimates the speed of the unmanned ship, i.e., the system state and the uncertain dynamic unknowns of the system, by establishing an extended state observer, and uses the output of this module as the input of other modules in the system;
[0043] The disturbance observer module is used to estimate the unknown dynamics of the unmanned ship at the next moment. Speed estimation of unmanned ship Calculate the total disturbance estimate of the system at the next moment
[0044] Specifically, the disturbance observer module is used to estimate the total disturbance to the system and use the output as the input to other modules;
[0045] The data-driven parameter update law module is used to estimate the total disturbance of the system at the next moment. Calculate the control gain estimate for the next moment
[0046] Specifically, the data-driven parameter update law module processes historical data online to estimate the variable control gain, and uses the output of this module as the input of other modules in the system.
[0047] Specifically, the system of the present invention can be established through the following steps:
[0048] Step 1: First, establish the unmanned ship dynamics model to obtain the unmanned ship speed at the next moment, including the dynamic unknown term of the inaccurate error in the control gain estimation;
[0049] Step 2: Establish a model-free controller module to generate a forward torque to adjust the speed of the unmanned ship;
[0050] Step 3: Establish a speed estimation module to estimate the speed of the unmanned ship, that is, the system state and the uncertain dynamic unknown items to which the system is subjected;
[0051] Step 4: Establish a disturbance observer module to estimate the total disturbance to the system;
[0052] Step 5: Establish a data-driven parameter update rate module to process historical data and estimate the variable control gain;
[0053] Step 6: Through the calculations of the previous modules, the system can estimate not only the control gain but also the inaccurate part of the control gain estimation based on the total disturbance of the system, thus realizing the function of full self-learning and monitoring.
[0054] Specifically, Figure 2 As shown, first, the model-free controller generates a forward torque control law through the difference between the speed of the unmanned ship and the expected speed, as well as the internal and external disturbances to the system. Then, the total disturbance estimate is expanded into an estimated state by establishing an extended state observer, which is input into the disturbance observer to track the total disturbance of the system by introducing an adaptive gain. Finally, the total disturbance estimate of the system is output to the data-driven parameter update rate module for historical data processing, so that the system effectively improves the tracking response speed of the control gain.
[0055] Preferably, the speed estimation module is:
[0056]
[0057] in, is the estimated speed of the unmanned ship at the current moment, is the estimated speed of the unmanned ship at the next moment, is the estimate of the unknown dynamics at the current moment, is the estimate of the unknown dynamics of the unmanned ship at the next moment, is the total disturbance estimate of the system at the current moment, λ 1 and λ 2 To adjust the expanded state error The introduced scalar gain parameter, T is the discrete time constant, and τ(n) is the forward torque output to push the ship forward.
[0058] Preferably, the disturbance observer module is:
[0059]
[0060] Among them, γ 1 is the adaptive gain, is the total disturbance estimate of the system at the next moment, is the total disturbance estimate of the system at the current moment, u(n) represents the speed of the unmanned ship at the current moment, u(n+1) represents the speed of the unmanned ship at the next moment, u(n)∈R, u(n+1)∈R, T is the discrete time constant, is the total disturbance estimate of the system at the current moment, is the control gain estimate at the current moment, and τ(n) is the forward torque output to push the ship forward.
[0061] Preferably, the calculation formula of the unmanned ship speed u(n+1) at the next moment is:
[0062] u(n+1)=u(n)+T(d+k 0 τ(n)) (3)
[0063] Among them, u(n) represents the speed of the unmanned ship at the current moment, u(n+1) represents the speed of the unmanned ship at the next moment, u(n)∈R, u(n+1)∈R; T is the discrete time constant, is the control gain, m is the mass parameter of the unmanned ship, τ(n) is the forward torque output to push the ship forward, and d is the dynamic unknown term.
[0064] Preferably, the calculation formula for the kinetic unknown term d is:
[0065] d=m -1 X u u(n)+m -1 X u|u| u(n)|u(n)|+m -1 ω (4)
[0066] Among them, d is the unknown term of dynamics, m is the mass parameter of the unmanned ship, and X u With X u|u| is the hydrodynamic parameter, u(n) represents the speed of the unmanned ship at the current moment, u(n)∈R; ω is the external disturbance suffered by the unmanned ship.
[0067] Preferably, the model-free controller module is:
[0068]
[0069] Among them, ω c is the model-free control gain, τ(n) is the forward torque output to push the ship forward, u(n) represents the speed of the unmanned ship at the current moment, u(n)∈R; is the control gain estimate at the current moment, is the total disturbance estimate of the system at the current moment, Estimate the unknown dynamics at the current moment.
[0070] Preferably, the data-driven parameter update law module is:
[0071]
[0072] in, is the control gain estimate for the next moment, γ 2 and γ 3 are the adaptive gains of the data-driven parameter update law module, is the control gain estimate for the next moment; τ(n) is the forward torque output to push the ship forward, τ T (n) is the inverse matrix of the forward torque that outputs the ship forward; u(n) represents the speed of the unmanned ship at the current moment, u(n)∈R; u(n+1) represents the speed of the unmanned ship at the next moment, u(n)∈R, u(n+1)∈R; T is the discrete time constant, is the control gain estimate at the current moment, is the total disturbance estimate of the system at the current moment.
[0073] Specifically, the value of the control gain is limited: The maximum amplitude is a_{max}
[0074] In this embodiment, the following parameters are introduced into the system of the present invention:
[0075] X u =-0.7225,X u|u| =-1.3274, ω = 0.1sin(t), when t≤50s, m = 47.6kg, t>50s, m = 23.8kg; λ 1 =10,λ 2 =50, T = 0.02s; γ 1 =4.25,ω c =25;γ 2 =0.01,γ 3 =0.0009, a_{max}=0.1, T=0.02s.
[0076] Using the above parameters, the system of the present invention is run:
[0077] like Figure 3 As shown in the figure, it is the control effect diagram of the unmanned ship speed, i.e. the control input. In the figure, u represents the speed of the unmanned ship, and u r By introducing the expected speed as a reference signal, the speed of the unmanned ship u can be shown to be proportional to the time-varying expected speed u r Conduct effective tracking.
[0078] like Figure 4As shown, this is the observation effect diagram of the total disturbance δ(n) of the system. In the figure, δ(n) represents the total disturbance of the system. It indicates the total disturbance estimation of the system. It can be shown that the total disturbance of the system is estimated effectively.
[0079] like Figure 5 As shown, it is a given signal diagram of the forward torque τ. It can be shown that when the speed of the unmanned ship changes with time, the system controller input that drives the unmanned ship to move forward is bounded.
[0080] like Figure 6 As shown, it is the control gain k 0 Observation effect diagram, in the figure k 0 represents the control gain, Indicates the control gain estimation. It can be shown that when the mass of the unmanned ship changes in 50 seconds, the system effectively estimates the changing control gain.
[0081] From the simulation results, it can be seen that the proposed discrete-time data-driven self-learning unmanned ship speed controller, by introducing an extended state observer, not only estimates the total disturbance, but also estimates the control gain. By introducing the data-driven self-learning method and effectively using historical data and current real-time data for calculation, it is possible to estimate the inaccurate part of the control gain in the total disturbance, and effectively improve the tracking response speed of the control gain estimation.
[0082] Beneficial results:
[0083] 1. In the existing unmanned ship speed self-disturbance control method, when estimating uncertain time-varying disturbances, the system needs to assume that the control gain is known or its nominal value is known. However, measuring the control gain requires a large amount of experimental data, which is not conducive to the engineering application of the method. The present invention, by introducing an extended state observer, can not only estimate the total disturbance of the ship, but also estimate the control gain at the same time.
[0084] 2. In the existing unmanned ship speed self-resistance control method, only the given control gain can be estimated, but the variable control gain cannot be processed. However, the present invention, by introducing a self-learning method, realizes the overall estimation of the total disturbance of the system including unknown environmental disturbances, model uncertainty and inaccurate estimation of the variable control gain.
[0085] 3. In the existing unmanned ship speed adaptive control method, historical data is not used, resulting in a slow convergence speed of adaptive parameters. However, the present invention uses a data-driven self-learning method to design a data-driven parameter update law, which can use the stored historical experience data to achieve tracking response to the control gain in a shorter time.
[0086] 4. To address the deficiencies of the prior art methods, the present invention combines the data-driven self-learning method with the extended state observer, and proposes a discrete-time data-driven self-learning unmanned ship speed controller with fast response speed, high efficiency and strong applicability. This design converts the system output and external disturbance into an extended state, and uses the extended state observer in the self-disturbance rejection technology to estimate the system state. Not only can the complex unknown time-varying total disturbance be estimated, but the control gain can also be estimated in a relatively short time.
[0087] 5. The purpose of the present invention is to estimate the total disturbance, by utilizing the extended state observer and the data-driven update law, not only to solve the problem of estimating the variable control gain, but also to identify the inaccurate part of the control gain estimation, and finally to achieve the effect of the system's complete self-learning and monitoring functions.
[0088] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.
Claims
1. A discrete-time data-driven self-learning unmanned ship speed control system. It is characterized in that include: Speed estimation module, disturbance observer module, model-free controller module, data-driven parameter update law module; The model-free controller module is used to calculate the forward torque τ(n); The velocity estimation module is used to calculate the unknown dynamics of the unmanned ship at the next moment according to the forward torque τ(n). Estimation of the speed of the unmanned ship at the next moment The speed estimation module is: in, is the estimated speed of the unmanned ship at the current moment, is the estimated speed of the unmanned ship at the next moment, is the estimate of the unknown dynamics at the current moment, is the estimate of the unknown dynamics of the unmanned ship at the next moment, is the total disturbance estimate of the system at the current moment, λ 1 and λ 2 To adjust the expanded state error The introduced scalar gain parameter, T is the discrete time constant, τ(n) is the forward torque output to push the ship forward; The disturbance observer module is used to estimate the unknown dynamics of the unmanned ship at the next moment. Speed estimation of unmanned ship Calculate the total disturbance estimate of the system at the next moment The disturbance observer module is: Among them, γ 1 is the adaptive gain, is the total disturbance estimate of the system at the next moment, is the total disturbance estimate of the system at the current moment, u(n) represents the speed of the unmanned ship at the current moment, u(n+1) represents the speed of the unmanned ship at the next moment, u(n)∈R, u(n+1)∈R, T is the discrete time constant, is the control gain estimate at the current moment, The data-driven parameter update law module is used to estimate the total disturbance of the system at the next moment. Calculate the control gain estimate for the next moment The data driven parameter update law module is: in, is the control gain estimate for the next moment, γ 2 and γ 3 are the adaptive gains of the data-driven parameter update law module; τ(n) is the forward torque output to push the ship forward, τ T (n) is the inverse matrix of the forward torque that outputs the ship forward; u(n) represents the speed of the unmanned ship at the current moment, u(n)∈R; u(n+1) represents the speed of the unmanned ship at the next moment, u(n+1)∈R; T is the discrete time constant, is the control gain estimate at the current moment, is the total disturbance estimate of the system at the current moment.
2. According to claim 1, a discrete-time data-driven self-learning unmanned ship speed control system, It is characterized in that The calculation formula of the unmanned ship speed u(n+1) at the next moment is: u(n+1)=u(n)+T(d+k 0 τ(n)) (3) Among them, u(n) represents the speed of the unmanned ship at the current moment, u(n+1) represents the speed of the unmanned ship at the next moment, u(n)∈R, u(n+1)∈R; T is the discrete time constant, is the control gain, m is the mass parameter of the unmanned ship, τ(n) is the forward torque output to push the ship forward, and d is the dynamic unknown term.
3. According to claim 2, a discrete-time data-driven self-learning unmanned ship speed control system, It is characterized in that The calculation formula of the kinetic unknown term d is: d=m -1 X u u(n)+m -1 X uu u(n)|u(n)|+m -1 ω (4) Among them, d is the unknown term of dynamics, m is the mass parameter of the unmanned ship, and X u With X uu is the hydrodynamic parameter, u(n) represents the speed of the unmanned ship at the current moment, u(n)∈R; ω is the external disturbance suffered by the unmanned ship.
4. According to claim 1, a discrete-time data-driven self-learning unmanned ship speed control system, It is characterized in that The model-free controller module is: Among them, ω c is the model-free control gain, τ(n) is the forward torque output to push the ship forward, u(n) represents the speed of the unmanned ship at the current moment, u(n)∈R; is the control gain estimate at the current moment, is the total disturbance estimate of the system at the current moment, is the estimate of the unknown dynamics at the current moment; u r Indicates the expected speed.
Citation Information
Patent Citations
A dynamic optimization controller structure of an unmanned marine vehicle and a design method
CN109583144A
Unmanned vessel data driving reinforcement learning control method with specified performance
CN111308890A
